External publications
From the ecosystem
External publications
Deep dives, benchmarks, and deployment write-ups about Dynamo, published by the customers and partners running it.
Featured
CoreWeave Forge: Turn AI Iteration Into Compounding ImprovementCoreWeave's platform for turning model iteration into compounding gains, with RL rollouts built on the same Dynamo foundation as CoreWeave Inference.Read on coreweave.comSWE-1.7: Frontier Intelligence at a Fraction of the CostHow Cognition trained SWE-1.7 with RL across clusters on three continents, using Dynamo to manage inference engine lifecycles and reroute work when a replica fails.Read on cognition.comPrime Inference: Fast, Reliable Serving for Frontier Open ModelsHow Prime Intellect serves frontier open models in production, with Dynamo routing and orchestrating disaggregated prefill and decode on top of vLLM.Read on primeintellect.ai
All publications
PyTorchOct 8, 2026Session-Aware Agentic Inference with NVIDIA DynamoViettel CloudOct 4, 2026Turning GPU-Hours Into Tokens: How Viettel Cloud Serves Multiple LLM Families With NVIDIA DynamoPrime IntellectOct 2, 2026Prime Inference: Fast, Reliable Serving for Frontier Open ModelsCoreWeaveSep 30, 2026CoreWeave Forge: Turn AI Iteration Into Compounding ImprovementHao AI Lab / UCSDSep 29, 2026FastVideo UniServe: Production Serving for FastH3 on NVIDIA Blackwell GPUs at 1.8x Lower Latency and 1.45x Higher ThroughputvLLMSep 29, 2026Taking vLLM Apart: A Practical Guide to Disaggregated ServingRed HatSep 29, 2026Red Hat Delivers Peak Performance on Kubernetes and CPUs in MLPerf Inference v6.1CrusoeSep 16, 2026Crusoe's MLPerf Inference v6.1 Results on NVIDIA GB200 NVL72NeuRealitySep 15, 2026One Model, Two GPU Vendors: Disaggregated Inference from H100 to MI300XPinterestSep 14, 2026Pinterest Builds a New Foundation for Multimodal AI With NVIDIAPinterestSep 11, 2026Building Pinterest's VLM Serving Stack on NVIDIA DynamoBasetenSep 11, 2026DeepSeek-V4.1-Flash: More Efficient Prefill for Coding Agents
